Artificial intelligence is a hungry beast. It subsists on a continuous input of trillions of data points, incessantly churning, chewing and spitting out insights. The data scientist or AI specialist must continuously seek new data sources to feed the beast and fine-tune their creation.
At Quandl we’re accustomed to providing financial professionals with new and unique data feeds to augment their trading strategies. In recent months, however, we’ve seen a notable increase in the number of AI and machine learning companies requesting such data. To us, this signals growth in the number of AI practitioners deploying functional algorithms in business. It’s clear that AI teams need more than the limited data they can scrape from the web.
But at the same time, it can be difficult to separate the wheat from the chaff.
Quandl has over 275,000 users on the platform and we pride ourselves on rigorously vetting, cleaning and testing the data we include on our platform. To that end, we are thrilled – but not surprised – with the uptick in AI users on the platform. While we don’t know exactly how our data is being incorporated into AIs (hey – this is the secret sauce of our users; we don’t ask!), we do know our customers who self-identify as AI/ML practitioners are actively using the below data feeds.
1. Global Economic Indicators: Trading Economics
Macroeconomic forecasting is typically the domain of the brightest minds in the economic world. Predicting a country’s potential GDP, its unemployment rates, property prices and public finance, however, requires countless man hours and skill. Luckily, as Dataconomy suggests, modeling tools and machine learning may aid in economic forecasting and bring new insight to policymaking.
Data scientists focused on this area of research have used Quandl’s Trading Economics data feed to build algorithms to uncover new patterns and signals. This data includes comprehensive, harmonized macroeconomic statistics for 200+ countries. 7,000+ indicators (GDP, CPI, PPI, IP, BOP, FDI, employment, etc.) from 1,000+ sources, updated hourly. Single user licenses start at USD 210/month.
2. Stock Market Algorithms: End of Day Stock Prices
AI and finance are made for one another. The massive volume of financial data is prime for machine analysis that can identify and even predict trading signals that a human may overlook. Like any opportunity, though, AI has the potential of being a threat to late and non-adopters. As VentureBeat predicts, “wealth managers who want to remain active in the investing process will be able to augment their decision making with new AI-driven tools and analysis.” While we don’t believe humans are going away anytime soon, we do see AIs supplementing human research in meaningful ways.
Key to this machine uprising? End of Day Stock Price data, which is popular among AI companies looking to develop stock market stimulators and predictions. Quandl offers professional-grade EOD stock prices, dividends, adjustments and splits for publicly traded US stocks. These are updated daily with history from 1996. Single user licenses start at USD 210/quarter.
3. Company Fundamentals: Zacks and Mergent
Alongside EOD stock price data, fundamentals data is equally important in training supervised learning algo-trading models. Zacks Fundamentals offers the most comprehensive coverage of stocks and indicators. On the Quandl platform you will find Zacks Fundamentals Collections A, B and C, all of which cover 17,000+ companies, with licenses starting at USD 1,200/year.
Mergent Global Fundamentals data is also a particularly popular dataset among our AI clients looking for fundamental indicators. It covers 49 fundamental indicators for 10,000+ companies from 67 countries as listed in the Russell Global Index. Single user licenses start at USD 185/month. Zacks Fundamentals would be the preferred option for US stocks, while Mergent offers greater global coverage.
4. Social Media Analytics and FinsentS for Sentiment Data
For many ML companies, data cleansing and classification is the most tedious task in their product development. Often AI companies will outsource this grunt work, but in some instances, they can buy data that has already been classified and tagged properly. Such is the case for social media and web sentiment data.
ML companies looking to build algorithms around sentiment data (opinion mining) frequently purchase Quandl’s FinSentS Web News Sentiment data feed. This product offers daily news sentiment indicators for 30,000+ stocks since 2013, derived from publicly available internet sources. Single user licenses begin at USD 50/month. Quandl also offers social media sentiment data, which is available upon request. Contact us for more details on this.
5. BNC for Cryptocurrency
Bitcoin is seeing a swell of machine learning methods focused on predicting the movements of cryptocurrencies. Given the hype and demand surrounding cryptos in the past year, it goes without saying that we’ve seen an increase in licenses for our BNC Liquid Index (BNC1) and BNC Digital Currency Indexed EOD (BNC2), which contain the first true historical price for bitcoin. Single user licenses for BNC1 begin from USD 100/month and BNC2 begin from USD 50/month.
For AI and ML, data reliability and accuracy is of the utmost importance. These user types need a trustworthy provider on which they can build their product. They need an efficient consumption experience and for data to be cleaned with the same diligence they would take themselves. Quandl meets these requirements. We take the tremendous pride in sourcing the highest quality dataset and ensuring a reliable delivery. You can read more about our process here or contact our sales team to learn more about our datasets.